The Euro and the Intensive and Extensive Margins of Trade ...

No 2008 – 06April

The Euro and the Intensive and ExtensiveMargins of Trade: Evidence from French

Firm Level Data_____________

Antoine BerthouLionel Fontagné

LBoivin

Texte tapé à la machine

The authors gratefully acknowledge ˝nancial assistance from the European Firms in a Global Economy: Internal policies for external competitiveness (EFIGE), a collaborative project funded by the European Commission's Seventh Framework Programme (contract number 225551).

The Euro and the Intensive and ExtensiveMargins of Trade: Evidence from French Firm

Level Data

Antoine BerthouLionel Fontagné

No 2008-06April

The Euro and the Intensive and Extensive Margins of Trade

TABLE OF CONTENTS

Non-Technical Summary 4

Abstract 6

Résumé Non-Technique 7

Résumé Court 9

1. Introduction 10

2. A First Glance at the Data 122.1. Data sources and construction . . . . . . . . . . . . . . . . . . . . 122.2. What the Data Say . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3. Data and Empirical Strategy 213.1. Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . 223.2. Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4. Estimation Results 254.1. Main Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254.2. Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

5. Conclusion 32

References 34

Appendix 35

3

CEPII, Working Paper No 2008-06

THE EURO AND THE INTENSIVE AND EXTENSIVE MARGINS OF TRADE: EVIDENCEFROM FRENCH FIRM LEVEL DATA

NON-TECHNICAL SUMMARY

Past literature on Currency Unions has found a positive but low effect of the adoption of acommon currency on trade flows within the Monetary Union. Beyond the macro perspectivehowever, very little has been done to investigate the effect of euro adoption at the micro-level.

On a theoretical perspective, previous literature suggests that the effect of a reduction infirm-level uncertainty related to nominal exchange rate volatility should lead to an increasein the average value of exports by each firm, and promote the entry of new exporters. Theadoption of the euro has also been motivated by the elimination of trade costs related toeconomic exchanges in different currencies, and by a greater price transparency within theeurozone. In a framework a la Melitz with heterogenous firms, the decrease in variable tradecosts would lead to an increase in the value of exports by firm in the foreign market, and tothe entry of new firms on the export market. A decrease in the fixed cost of export howeverwould only have a positive effect on the number of exporting firms. More recently, modelsof trade with multi-product firms have emphasized that a variation of trade costs may lead toa within-firm adjustment through the number of goods exported by each firm. Finally, onewould expect from a greater price transparency that competition increases, thus contributingto firm selection on the export market. Hence, theory tells us that the effect of the euro ontrade can both translates into a variation in the number of exporting firms or product exportedby firm, and a variation in the value of exports by firm or by product.

We contribute to the literature on the micro effects of Currency Unions, by making use ofa unique database on French firm-level exports. Previous empirical studies on the micro ef-fects of the euro have been using product-level trade data at the HS6 level - with around 5,000product categories. However, the level of disaggregation of these data does not enable to con-trol for the entry or exit behavior of firms within a product category. In contrast, we rely onthe use of French data on firm-level exports provided by the French customs. Importantly,our data provide information on firm exports with detailed information on the nature of theproducts that are shipped (HS8 categories). We can therefore identify an additional patternrelated to the fact that several firms can export within a single product category, and we definea variety as a product category exported by a single firm. Note that we restrict our analysis tomanufactured products, including agro-food. Our data cover the 1998-2003 period.

Since we are interested in measuring the effect of the euro on trade, we identify a "treat-ment group", i.e. destination countries located within the eurozone (EZ13: eurozone 13 less

4


Greece and Slovenia), and a "control group" in which we identify three destination regions:the rest of the EU15 less Greece (nonEZeu); the rest of Europe including the 12 enlargementcountries (less Malta and Cyprus) plus Switzerland and Norway (nonEZeurope); the rest ofthe world (nonEZworld).

We define the extensive margin as the number of varieties that are exported to each desti-nation country, and the intensive margin as the average value of exports by variety. Theseexport margins have a bilateral dimension, and are used in the econometric part of the paper.In the descriptive statistics, we further identify the intensive and extensive margins by des-tination region, which requires to account for the geographical component of the extensivemargin, i.e. the number of destination countries for each variety within a destination region.This second decomposition of the export margins has therefore a multilateral dimension.

Relying on the latter definition, our descriptive statistics indicate that overall, the variation ofexports to eurozone partners over the period is proportionally more driven by the extensivemargin. This is the result of different changes. The number of French exporters has declinedin the eurozone, contributing negatively to the extensive margin of trade. In contrast, the av-erage number of varieties exported by firm, and the average number of destination markets byvariety - within the EZ destination region - have both positively contributed to the extensivemargin. Finally, the intensive margin shows a positive trend over the period. We record asimilar increase in the number of products exported by firm to the eurozone and to nonEZeudestinations; however the reduction in the number of exporters is more pronounced in theeurozone than in nonEZeu, which dampens the increase in the extensive margin. Exportstowards the rest of Europe are characterized by a decrease in the number of varieties exportedby firm, a slight increase in the number of exporting firms and destination markets by variety,and a very large increase in the intensive margin. This pattern is similar and even amplifiedfor exports to the rest of the world, with a very large decrease in the number of varietiesexported to nonEZworld partners. Overall, the descriptive statistics indicate that if any, theeuro had a differentiated impact on the two margins of trade.

As a second step we use the decomposition of the intensive and the extensive margins usingthe bilateral dimension: the extensive margin is therefore defined as the number of varietiesexported to each partner, and does not include the geographical component as in the descrip-tive statistics. We proceed to econometric estimates that aim at identifying the effect of theeuro on French exports to EZ partners, as compared to French exports to nonEZ partners. Weestimate two gravity equations with the intensive and extensive margins of French exports asthe dependent variables. In our empirical methodology, we introduce a measure of the nom-inal exchange rate volatility, in order to control for the reduction in the related uncertaintyat the firm level. We also introduce eurozone dummies to control for additional effects that

5


the adoption of the euro may have generated on French exports, as suggested by the literature.

Estimation results point to a strong negative effect of the nominal exchange rate volatilityon both the intensive and extensive margins of trade, which suggests a large - positive - ef-fect of the euro related to the reduction of firm-level uncertainty. Volatility indeed reducesthe number of varieties that are shipped to trade partners, as well as the value of exports byvariety.

The coefficients on the euro dummies also indicate an additional positive effect of the euro onthe extensive margin, which is not related to the reduction in nominal exchange rate volatil-ity. We find that the difference between the number of varieties exported to EZ partners andnon-EZ partners, has increased by 19.4% between 1998 and 2003. We also find a positiveeffect specific to the introduction of the banknotes, that can be estimated at 7.8%. Robust-ness checks indicate however that the positive effect of the euro introduction on the extensivemargin of French exports is responsive to the size of the control group.

Overall, this paper contributes to the literature on the micro effects of Currency Unions,and provides new evidence of a positive effect on trade that is not related to the eliminationof nominal exchange rate volatility. This effect can be attributed to reduced trade costs orincreased price transparency, which has increased the number of varieties exported by Frenchfirms.

ABSTRACT

We improve the study of the effects of a Currency Union on trade. Using data on Frenchexports at the firm level, we compute an intensive and extensive margins of French exports- with a variety dimension - over the period 1998-2003. Estimation results indicate thatnominal exchange rate volatility has a negative effect, which translates into the intensive andextensive margins. We also provide some evidence that the euro had an additional positiveeffect on the extensive margin; this effect is not related to the reduced nominal exchange ratevolatility. This suggests a new varieties effect of the euro.

JEL Classification: F15Keywords: trade, export margins, euro.

6


EFFET DE L’EURO SUR LES MARGES INTENSIVE ET EXTENSIVE DU COMMERCE :APPROCHE PAR L’UTILISATION DE DONNÉES DE FIRMES FRANÇAISES

RESUME NON TECHNIQUE

La littérature sur les Unions monétaires fait généralement état d’un effet macro-économiquepositif mais limité de l’adoption d’une monnaie commune sur les échanges commerciaux ausein de ces Unions. La question des effets micro-économiques de l’adoption d’une monnaieunique sur les échanges a néanmoins reçu jusqu’ici une attention limitée.

Théoriquement, la réduction de l’incertitude au niveau de la firme liée à la volatilité du tauxde change peut entraîner une augmentation des exportations de chaque firme, mais aussi fa-voriser l’entrée de nouveaux exportateurs. L’adoption de l’euro a également été motivée parune élimination des coûts de commerce liés à l’existence d’échanges dans des monnaies diffé-rentes, et par une hausse de la transparence des prix. Dans un contexte de firmes hétérogènesa la Melitz, la réduction des coûts de commerce variables doit entraîner à la fois une haussedes ventes par firme à l’étranger ainsi qu’une augmentation du nombre de firmes exporta-trices, alors qu’une baisse du coût fixe ne joue positivement que sur le nombre de firmesexportatrices. Plus récemment, les modèles avec firmes multi-produits ont mis en évidencele fait que l’ajustement consécutif à une libéralisation commerciale peut se faire au sein dela firme, par le nombre de biens produits et exportés. Enfin, on peut attendre d’une haussede la transparence sur les prix, une hausse de la concurrence au sein de l’Union Monétaire,contribuant à la sélection des firmes sur le marché d’exportation. Par conséquent, les effetsde l’euro sur le commerce peuvent se traduire à la fois par une hausse du nombre de firmesexportatrices ou par le nombre de produits exportés par firme, ainsi que par une hausse duvolume d’exportation par firme ou par produit.

Nous nous proposons ici d’analyser, à l’aide de données individuelles de firmes, la ma-nière dont l’introduction de la monnaie unique a affecté le nombre de firmes exportatrices, lenombre de variétés exportées, ainsi que le volume d’exportation par variété. Les études empi-riques des impacts micro-économiques de l’euro ont jusqu’ici utilisé des données d’échangescommerciaux au niveau de la nomenclature HS6 (comprenant 5 000 produits environ). Tou-tefois, même à ce niveau de désagrégation, il est impossible de contrôler pour l’entrée ou lasortie de firmes exportatrices au sein de chaque catégorie de produit. A cette fin, nous mobili-sons des données individuelles de firmes françaises, mises à notre disposition par la Directiondes Douanes.

Nous disposons pour chaque année sur la période 1998-2003 du montant des exportationspour chaque firme présente sur le marché d’exportation, détaillé par catégorie de produit auniveau HS8. Ce détail dans les données nous permet ainsi d’observer le nombre de firmesexportant au sein d’une même catégorie de produits, ce qui n’est bien entendu possible que sil’on dispose de données de firmes. Nous considérons ici que deux firmes exportant un mêmeproduit exportent en réalité deux variétés différentes. Nous limitons notre analyse aux pro-duits manufacturés, incluant l’agro-alimentaire.

7


Afin d’identifier les effets de l’adoption de l’euro sur les échanges commerciaux, nous dé-finissons un "groupe de traitement", à savoir les pays de la zone euro (EZ13 : la zone euromoins la Grèce et la Slovénie), ainsi qu’un "groupe de contrôle" distinguant trois régions dedestination : le reste de l’UE15 à l’exception de la Grèce (nonEZeu), le reste de l’Europeincluant les 12 pays de l’élargissement (moins Chypre et Malte) ainsi que la Suisse et la Nor-vège (nonEZeurope) ; enfin le reste du monde (nonEZworld).

Nous définissons la marge extensive du commerce comme le nombre de variétés qui sontexportées vers chaque destination, et la marge intensive comme la valeur moyenne des ex-portations par variété. Ces marges d’exportation ont une dimension bilatérale, qui est mobi-lisée dans la section économétrique de cet article. Dans la section proposant des statistiquesdescriptives, nous ajoutons une composante géographique à la marge extensive, en prenanten compte la diversité des marchés d’exportation au sein d’une région de destination, pourchaque variété.

Nos statistiques descriptives mettent en évidence que globalement, la croissance de la va-leur des exportations manufacturières totales de la France à destination des partenaires dela zone euro sur la période considérée s’explique davantage par la marge extensive du com-merce que par la marge intensive. Concernant la marge extensive, ce résultat s’explique par lacombinaison de plusieurs évolutions : le nombre d’exportateurs français vers la zone euro adécliné, contribuant à réduire la marge extensive ; mais le nombre moyen de produits exportéspar chaque firme a augmenté, et donc joué en sens opposé ; enfin, le nombre de destinationspour chaque variété a lui aussi augmenté. Nous notons également une contribution positivede la marge intensive sur la période. Comparant les régions de destination entre elles, nousobservons une augmentation similaire du nombre de produits exportés par firme vers la zoneeuro et vers les destinations nonEZeu. Toutefois, le recul du nombre d’exportateurs est plusprononcé au sein de la zone euro qu’au sein de la destination nonEZeu, ce qui limite la contri-bution de la marge extensive au sein de la zone euro. Les exportations à destination du restede l’Europe se caractérisent par une baisse du nombre de produits exportés par firme, une lé-gère augmentation du nombre de firmes exportatrices et du nombre de marchés de destinationpar variété. Enfin la croissance de la marge intensive y est très prononcée. Nous observonsdes évolutions similaires, mais encore plus marquées, dans le cas des exportations françaisesvers le reste de monde, notamment un recul très net du nombre de produits exportés par firme.Ainsi, ces statistiques descriptives suggèrent-elles que si l’euro a joué un rôle, son impact aété très différent sur les deux marges du commerce.

Dans une seconde étape, nous utilisons la décomposition des marges intensive et extensive enexploitant la dimension bilatérale de celles-ci, comme indiqué plus haut. La marge extensiveest alors définie comme le nombre de variétés exportées vers chaque destination, et ne prendpas en compte la composante géographique utilisée dans les statistiques descriptives. Nousproposons une estimation économétrique visant à identifier les effets de l’introduction de lamonnaie unique sur les exportations françaises vers les destinations de la zone euro, par com-paraison avec les exportations vers les autres destinations. Nous estimons deux équations de

8


gravité utilisant comme variable dépendante chacune des deux marges du commerce. Nousutilisons notamment comme variable explicative la volatilité du taux de change nominal, afind’identifier l’effet d’une réduction de l’incertitude au niveau de la firme. Nous introduisonségalement dans les variables explicatives des variables indicatrices zone euro, pour mesurerles effets additionnels de l’introduction de l’euro sur les exportations françaises, comme sug-géré par la littérature.

Les résultats d’estimation indiquent que la volatilité du taux de change nominal a un effetnégatif, à la fois sur les marges intensive et extensive du commerce, ce qui suggère un effetpositif de l’euro lié à la réduction de l’incertitude au niveau de la firme. La volatilité du tauxde change nominal réduit en effet à la fois le nombre de variétés exportées, ainsi que le vo-lume moyen d’export par variété.

Les coefficients sur les variables indicatrices zone euro indiquent également un effet addi-tionnel positif de l’euro sur la marge extensive, qui n’est pas lié à la réduction de la volatilitédu taux de change. La différence entre le nombre de variétés exportées vers des partenaireszones euro et hors zone euro s’est accrue de 19,4% sur la période. L’effet spécifique de l’intro-duction des billets en euro sur les exportations françaises n’est pas négligeable. La différenceentre le nombre de variétés exportées vers les partenaires de la zone euro et vers les autrespays a augmenté d’environ 8,7% suite à l’introduction des billets. Les tests de robustesseeffectués soulignent néanmoins que l’impact positif de l’introduction de l’euro sur la margeextensive des exportations françaises est sensible à la taille du groupe de contrôle.

Plus généralement, cet article contribue à la littérature sur les effets micro-économiques desunions monétaires, et confirme l’existence d’un effet positif sur le commerce, au-delà del’élimination de la volatilité du taux de change. Cet effet, qui peut être interprété comme laconséquence d’une baisse des coûts de commerce ou encore à une hausse de la transparencesur les prix, a contribué à accroître le nombre de produits ou variétés exportés par chaquefirme.

RESUME COURT

Cet article contribue à l’étude des effets de l’établissement d’une Union Monétaire sur lecommerce. Nous utilisons des données d’exportation au niveau de la firme pour construireles marges intensive et extensive des exportations françaises sur la période 1998-2003. Lesrésultats d’estimation indiquent que la volatilité du taux de change nominal a un effet négatifsur les deux marges du commerce. Au-delà de l’élimination de la volatilité du taux de changenominal, nous trouvons que l’adoption de l’euro a eu un effet supplémentaire positif sur lamarge extensive. Ces résultats suggèrent donc un effet positif de l’euro sur le nombre devariétés exportées par les firmes françaises.

Classification JEL : F15Mots clés : Commerce, marges du commerce, euro.

9


THE EURO AND THE INTENSIVE AND EXTENSIVE MARGINS OFTRADE

Antoine BERTHOU1, Lionel FONTAGNÉ2

1. INTRODUCTION

Past literature on Currency Unions has found a positive effect of the adoption of acommon currency on trade flows within the Monetary Union. Rose (2000) finds thatcommon currency raises trade by 235%. Estimates on the effect of the euro on tradeare however typically lower. Baldwin (2006) reviews the literature and acknowledgesthat the effect should actually be between 5% and 10%. Beyond the macro perspec-tive however, very little has been done to investigate the effect of euro adoption at themicro-level.

In a theoretical perspective, Baldwin and Taglioni (2004) develop a Melitz-like trademodel where each firm exports a single good, and find that reduced uncertainty re-lated to nominal exchange rate volatility enables exporters to export larger volumes,as well as the entry of new - smaller - exporters. Beyond reduced volatility, the eurohas also been motivated by the elimination of trade costs related to currency exchangeand related transaction costs. In a Melitz (2003) framework with heterogenous firms,a decrease in the fixed entry cost would lead to the entry of new - less productive andsmaller - firms into the export market. In a similar framework, a decrease in the vari-able trade cost contributes to firm selection, but also increases the value of exports byvariety. We may therefore expect that the adoption of the euro has contributed at theaggregate level to the export of new varieties to each eurozone partner, and to someextent to the increase in the value of exports by firm, depending on how the euromay have affected trade costs. More recently, Bernard et al. (2006) have developeda model of multi-product firms. In their model, a decrease in the variable trade costinduced by trade liberalization leads to firm selection into the export market, but alsoincreases the number of goods that are exported by each firm3.

1Panthéon-Sorbonne-Economie, Université Paris I, Paris School of Economics. Email:[email protected]

2Paris School of Economics, Université Paris I and CEPII. Email: [email protected] .We are indebted to participants to the HEI workshop ”Trade and the Euro” held on January 12th 2008in Geneva, for their helpful comments on a preliminary version of this paper. We also thank MatthieuCrozet for his advises on the database and his suggestions, and Agnès Bénassy-Quéré for her comments.

3The range of goods produced by each firm decreases while the proportion of goods that are exportedis larger.

10


Potentially, the adoption of the euro may have generated a positive effect on Frenchexports via a reduction in the volatility of the nominal exchange rate, but also viareduced fixed or variable trade costs, associated to the elimination of the cost of ex-changing currencies. The adoption of the euro may have also increased price trans-parency, leading to greater competition and enhanced firm or product selection.

Testing those micro mechanisms requires to use highly disaggregated data. Usingproduct-level trade data at the HS6 level - with around 5,000 product categories,Baldwin and Nino (2006) find some support for a New Goods hypothesis of the euro,while Flam and Nordstrom (2007) find some evidence that most of the effect of theeuro goes through the extensive margin of trade, defined as the value of exports inproducts that are not continuously traded over time. However, the level of disag-gregation of these data does not enable to investigate the effect of the euro on firmsbehavior.

In this paper, we use a unique database on French exports at the firm level, providedby the French customs. In our approach, we define a variety as a product categoryexported by a single firm. Using the French data, we build an intensive and exten-sive margins of French exports, that correspond respectively to the value of exportsby variety, and the number of varieties exported to each destination country. For thedescriptive statistics section of the paper, we also compute a multilateral extensivemargin, that integrates a geographical dimension4. Our data cover the 1998-2003 pe-riod for French exports. The descriptive statistics indicate that overall, the variationof exports to eurozone partners over the period is proportionally more driven by theextensive margin. The empirical part of the paper also provides some support to theNew Variety hypothesis of the euro. We find strong evidence that a reduced volatilityof nominal exchange rate contributes to the increase in the number of varieties ex-ported and to an increase in the value of exports by variety. Results also suggest thatthe euro had a positive effect on the number of varieties that are exported by Frenchfirms, consistent with a decrease in the fixed entry cost and enhanced competitionleading to further firm and product selection on the export market.

The remaining of the paper proceeds as follows: Section 2 describes the data andprovides a first piece of evidence using descriptive statistics; we develop our empiri-cal strategy in Section 3; estimation results are provided in Section 4; finally, Section5 concludes.

4Working with destination regions requires to take into account the number of destinations for eachvariety within a region.

11


2. A FIRST GLANCE AT THE DATA

2.1. Data sources and construction

We use individual data on French firms provided by the French Customs. Our datasetcomprises all custom records at the firm, product and destination country level for theperiod 1998-2003. Hence we have only one year before the introduction of the euro.However, the dataset is large enough to study the dynamics of the euro introductionon trade flows over the recent period. Most importantly, the database enables to dis-tinguish between the number of varieties that are exported to each destination market- the extensive margin, from the value of exports by variety - the intensive margin. Itis worth noticing that the use of firm level data has the great advantage to enable theconstruction of export margins with a variety dimension, which is not the case withproduct level databases.5 Within country pairs, we define the extensive margin witha variety - product × firm - dimension rather than with a simple firm dimension,since firms are likely to export several goods or product categories, as emphasized inBernard et al. (2007). We therefore fully exploit the dimension of the database pro-vided by the French customs. We come back later in this section on the constructionof the intensive and extensive margins, and give some details on the construction ofour database.

The database provided by the French customs classifies export flows at the firm levelwithin the product categories of the Combined - 8 digit - nomenclature (CN8). Werestrict our analysis to manufactured products, including agro-food. Accordingly, wedrop all product categories that do not match ISIC-rev2 headings 311 to 390.6 Thisleaves us with all export flow data within 28 manufacturing industries.

Using the French data, we provide two methodologies to decompose the intensiveand extensive margins of French exports, (i) within country pairs - export marginswith a bilateral dimension - and (ii) within destination regions - export margins witha multilateral dimension. Note that the multilateral decomposition will be used in thedescriptive statistics, while the bilateral decomposition will be used in the economet-ric part of the paper.

5For instance, the BACI (CEPII) or COMTRADE (UN) databases provide bilateral trade flow datawith HS 6-digit classification; these databases therefore only enable a count of product categories thatare exported at the bilateral level.

6We rely on the classification of HS6 positions - the first 6 digits of the CN8 - in ISIC headings,which is provided by the CEPII and is available on the CEPII website.

12


2.1.1. A decomposition of export margins within country pairs

We first detail how we compute the intensive and extensive margins of French exportsfor each destination country and industry. For this exercise, we use shipments fromFrance to each destination market for each one of the 28 ISIC rev.2 manufacturingindustries. We define the extensive margin of French exports to country j withinindustry k, as the number of product varieties exported to each destination country.We compute the extensive margin as follows:

Ekjt = Nkjt × Zkjt; with Zkjt =PNkjt

n=1 mnkjt

Nkjt

Ekjt is the extensive margin of exports to country j, in industry k, at time t; Nkjt isthe number of exporting firms indexed n = 1...Nkjt that export to country j; Zkjt isthe average number of product categories exported by firm to country j and mnkjt isthe actual number of product categories exported by each individual firm indexed nto the same destination.

We define the intensive margin of French exports as the average value of shipmentsby variety to each destination country. We compute the intensive margin as follows:

Ikjt =PEkjt

i=1 xikjt

Ekjt

Ikjt is the intensive margin of French exports to destination j, in industry k, at time t;xikjt is the value of shipment for variety i. This decomposition of the value of Frenchexports at the bilateral level is used in the econometric part of this paper.

2.1.2. A decomposition of export margins by destination region

Since we are interested in measuring the effect of the euro on trade, we identifya ”treatment group”, i.e. destination countries located inside the eurozone, and acontrol group in which we identify three destination regions. For the descriptivestatistics part of the analysis, we rely on the definition of four destination regioncategories:

• The eurozone 13 less Greece and Slovenia (EZ thereafter: Belgium and Lux-embourg7, Germany, Ireland, Spain, Italy, the Netherlands, Austria, Portugal,Finland);

7Belgium and Luxembourg are treated as a single country in our investigations, because of datarestrictions at the beginning of the period.

13


• The rest of the EU15 less Greece (nonEZeu)8;

• The rest of Europe including the 12 enlargement countries, less Cyprus andMalta, plus Switzerland and Norway (nonEZeurope)9;

• The rest of the world (nonEZworld).

We also compute the intensive and extensive margins of French exports to each oneof the four destination regions defined above, in order to build our descriptive statis-tics.10 Since we have several countries within each destination region, we need torefine the construction of the extensive margin to account for the fact that a single va-riety can be exported to several countries. We now use all the available information inorder to better identify all the components of the extensive margin (E): the extensivemargin can be defined as the product of the number of firms (N), the average numberof products exported by firm (Z), and the average number of destination markets foreach exported variety (G). As in the previous sub-section, the intensive margin (I) isdefined as the average value of exports for each variety exported to a given destinationj. The total value of exports (V) at time t can therefore be decomposed as followsfor each destination region R ∈ {EZ; nonEZeu; nonEZeurope; nonEZworld}:

V Rt = ER

t × IRt = NR

t × ZRt ×GR

t × IRt

Our decomposition includes all the dimensions of the multilateral extensive margin.In particular, we can isolate the geographical component (G) within our multilateralextensive margin.

2.2. What the Data Say

Before turning to the econometric part of the paper, we use the multilateral decom-position of the French export margins within destination regions, as defined above, toshed light on the effect of the euro on French export margins.

We start with the overall sample of sectors and destinations and decompose the totalvalue of exports into a number of exporting firms and an average value of exports

8Note that the Danish Krona is linked to the euro via the ERM II mechanism, with +/- 2.25%fluctuations possibilities.

9Estonian currency was pegged to the DM until 1999, and is now pegged to the euro.10Mauro Pisu and Laszlo Halpern, have replicated this methodology with Belgian and Hungarian

firm-level data respectively; their results have been used for the forthcoming 2008 EFIM report.

14


Tabl

e1:

Des

crip

tive

Stat

istic

s,al

ldes

tinat

ions

and

indu

stri

es19

9819

9920

0020

0120

0220

03To

talv

alue

ofex

port

s(b

illio

nseu

ros)

225

236

272

280

268

264

Mar

kets

hare

s(1

0%la

rges

texp

orte

rs)

0.94

0.94

0.95

0.95

0.95

0.95

Nb

ofex

port

ers

92,9

0894

,497

95,2

2795

,162

94,2

7091

,425

Ave

rage

valu

eby

expo

rter

(mill

ions

euro

s)2.

431

2.49

52.

854

2.93

92.

846

2.88

4A

vera

genb

ofsh

ipm

ents

byex

port

er18

.03

18.2

118

.68

18.7

518

.47

18.3

9A

vera

geva

lue

bysh

ipm

ent

134,

827

137,

020

152,

732

156,

756

154,

091

156,

843

15


by firm. We also compute the average number of shipments by firm11, as well asthe average value of export by shipment. Finally, we report the market share of the10% largest exporters12. All these statistics are provided in Table 1 for the period1998-2003. The first striking fact is the extreme concentration of exports. Only fewfirms export. The 10 percent ’champions’ concentrate 94 percent of the total valueof French exports in 1998 and even 95 percent in 2003. This reproduces the styl-ized fact stressed by Mayer and Ottaviano (2007). The total number of firms rathertends to decrease over the period. One percent and a half of the total number of ex-porters have disappeared over the considered period; this tendency can be explainedby mergers that occurred during the period13. We then compute the average numberof shipments by each exporting firms. This number varies over time in a range of 18to 19 shipments. Finally, results indicate that the average value of exports by firm,as well as the average value of exports by shipment, increases over time. Overall,French exports seem to have been driven both by the intensive and extensive marginsover the period, with less firms exporting more product categories to more destina-tions, together with a larger value of exports by shipment.

In Table 2, we replicate the same exercise for each of the four destination regionsdefined in the previous section. Not surprisingly, EZ is the first destination of exportsfor French firms, with some 50 percent of the value of the shipments throughout theconsidered period. The concentration of exports is larger for nonEZworld; this fea-ture can be related to higher fixed and variable costs of exports outside Europe, whichonly enable larger and more productive firms to enter. No clear difference in termsof concentration is observed within Europe across our three groups of destination.The value of exports by firm is much larger for the eurozone, and has been increasingquickly over the period that we consider. The number of shipments by firm is indeedlarger within the eurozone, and even dominates the number of shipments by firm tothe rest of the world. The value of exports by shipment is also large for exports to theEZ, but is dominated by the value of exports by shipment to the nonEZeu destinationregion.

In Table 3, we specifically use a decomposition of the total value of export flows intoan extensive and an intensive margins, while still reporting all the components of the

11Which can also be considered as the number of varieties exported by firm if one considers a productexported to a single destination country as a variety.

12In terms of export sales.13Observing mergers however requires to have information on the financial linkages of firms. This

question goes beyond the scope of our study.

16

The Euro and the Intensive and Extensive Margins of TradeTa

ble

2:D

escr

iptiv

est

atis

tics

(con

t’d)-

alli

ndus

trie

s,by

dest

inat

ion

Tota

lval

ueof

expo

rts

(bill

ions

)M

arke

tsha

res

(10%

larg

este

xpor

ters

)E

Zno

nEZ

euno

nEZ

Eur

ope

nonE

Zw

orld

EZ

nonE

Zeu

nonE

ZE

urop

eno

nEZ

wor

ld19

9811

428

.514

.568

.90.

900.

900.

900.

9419

9912

130

.315

.868

.50.

910.

900.

910.

9420

0013

733

.817

.383

.80.

910.

910.

910.

9520

0113

734

.718

.689

0.90

0.90

0.91

0.95

2002

132

34.4

18.5

83.3

0.90

0.90

0.91

0.94

2003

133

31.8

18.8

800.

910.

900.

910.

95

Nb

Exp

orte

rsV

alue

ofex

port

sby

expo

rter

(mill

ions

)E

Zno

nEZ

euno

nEZ

Eur

ope

nonE

Zw

orld

EZ

nonE

Zeu

nonE

ZE

urop

eno

nEZ

wor

ld19

9845

,388

22,2

7236

,156

62,8

542.

511

1.27

70.

402

1.09

719

9946

,390

22,9

2336

,824

63,7

072.

611

1.32

30.

429

1.07

520

0046

,571

23,1

2737

,642

64,7

332.

939

1.46

20.

459

1.29

420

0143

,317

22,2

5938

,413

66,8

493.

168

1.56

0.48

51.

332

2002

41,7

4921

,582

38,4

8966

,701

3.16

61.

594

0.48

1.24

820

0341

,561

21,3

7536

,988

63,9

353.

202

1.48

70.

508

1.25

2

Nb

ofsh

ipm

ents

byex

port

erV

alue

ofex

port

sby

ship

men

tE

Zno

nEZ

euno

nEZ

Eur

ope

nonE

Zw

orld

EZ

nonE

Zeu

nonE

ZE

urop

eno

nEZ

wor

ld19

9814

.92

6.63

4.72

10.8

116

8,24

719

2,59

385

,141

101,

437

1999

15.4

36.

954.

7910

.49

169,

190

190,

415

89,4

2210

2,41

120

0015

.63

7.05

4.9

10.8

818

8,09

520

7,44

193

,653

118,

993

2001

16.5

77.

264.

9210

.71

191,

251

214,

961

98,6

5212

4,38

520

0216

.91

7.24

4.87

10.3

718

7,21

422

0,27

398

,595

120,

364

2003

17.3

37.

314.

729.

8618

4,77

320

3,49

910

7,52

512

7,03

2

17


extensive margin. In 1998, we have 316,354 varieties - N × Z - that are exportedwithin the EZ region to 2.14 partners on average. This translates into an extensivemargin of 677,272 shipments within the region, with an average individual value of168,247 euros by shipment - i.e. the intensive margin. Overall, the product of theintensive and extensive margins gives a total value of exports by French firms withinthe EZ region of 113,95 billions euros that has been mentioned in the previous Table.Maybe the most important feature to highlight is that the data reveal that French ex-porters are typically multi-product firms, as suggested by Bernard et al. (2006) andBernard et al. (2007); each firm however exports its products to very few partners onaverage14.

We also report the 1998-2003 percentage changes in the components of the exten-sive margin, as well as the percentage change in the intensive margin, in Table 3and summarize the results in Figure 115. We observe that the number of French ex-porters has declined in the eurozone, contributing negatively to the extensive marginof trade16. In contrast, the average number of varieties exported by firm, and the av-erage number of destination markets by variety - within the EZ destination region -have both positively contributed to the extensive margin. Finally, the intensive mar-gin shows a positive trend over the period. Turning to the nonEZeu partners, thedescriptive statistics indicate that the pattern is similar, with slight differences. Whilethe decrease in the number of exporting firms is smaller than for the EZ destinationregion, there has been no variation over the period in the average number of destina-tion markets by variety. This can potentially be explained by the fact that the numberof destination countries in nonEZeu is lower, which limits the potential for exportingexisting varieties to new partners. The size of the destination region can thereforehave an influence on our results. Overall, we notice a similar increase in the numberof products exported to the eurozone and to nonEZeu; however the reduction in thenumber of exporters is more pronounced in the eurozone than in nonEZeu, whichnegatively impacts the extensive margin.

Exports towards the rest of Europe are characterized by a decrease in the numberof varieties exported by firm, a slight increase in the number of exporting firms and

14Note that this result may be due to the fact that many - small - firms only export to one partner,while few large firms, the ”happy few”, export large amounts to a number of foreign partners.

15Note that the negative variation in the number of firms over the period may be overestimated for theEU: the declaration threshold to the customs has indeed increased in 2001 and 2002, therefore leadingto artificial exits of firms exporting small amounts.

16Note again that this tendency may be associated to mergers that occurred intensively within the EUduring the period that we consider.

18


Tabl

e3:

Ext

ensi

vean

dIn

tens

ive

Mar

gins

ofTr

ade

-All

man

ufac

turi

ngin

dust

ries

,all

firm

s,by

dest

inat

ion

Lev

els

in19

98N

bex

port

ers

Nb

prod

ucts

byfir

mN

bde

stin

atio

nsby

vari

ety

Ext

ensi

veIn

tens

ive

NZ

GE

IE

Z45

,388

6.97

2.14

677,

272

168,

247

nonE

Zeu

22,2

724.

691.

4214

7,72

719

2,59

3no

nEZ

euro

pe36

,156

3.68

1.28

170,

617

85,1

41no

nEZ

wor

ld62

,854

5.76

1.88

679,

562

101,

437

Var

iatio

n19

98-2

003

(%va

riat

ion)

NZ

GE

IE

Z-0

.088

0.10

50.

044

0.06

10.

094

nonE

Zeu

-0.0

410.

096

00.

055

0.05

5no

nEZ

euro

pe0.

023

-0.0

410.

042

0.02

30.

233

nonE

Zw

orld

0.01

7-0

.105

0.01

3-0

.076

0.22

5

19


Figure 1: Intensive margin and the components of the extensive margins: percentchange 1998-2003

-15%

-10%

-5%

0%

5%

10%

15%

20%

25%

30%

35%

EZ11 nonEZeu nonEZeurope nonEZworld

IGZN

destination markets by variety, and a very large increase in the intensive margin. Thispattern is similar and even amplified for exports to the rest of the world, with a verylarge decrease in the number of varieties exported to nonEZworld partners.

We summarize the variations in the extensive and intensive margins in Figure 2. Fromthese descriptive statistics, it is obvious that the evaluation of the effect of the euroon the intensive and extensive margins of French exports margins highly depends onthe choice of the control group. Considering all destination regions, the data indicatethat French exports to EZ destinations have been proportionally more driven by anincrease in the number of varieties exported, which tends to confirm a new varietieseffect of the euro. If we restrict however the control group to the nonEZeu destinationregion, the picture is obviously less clear.

In the empirical part of the paper, our econometric methodology enables to isolate theeffect of the euro on the two margins of French exports, by controlling in particularfor factors such as EU membership that may generate endogeneity issues.

20


Figure 2: Extensive and Intensive Margins of trade: percent change 1998-2003

-10%

-5%

0%

5%

10%

15%

20%

25%

30%

EZ11 nonEZeu nonEZeurope nonEZworld

IE

3. DATA AND EMPIRICAL STRATEGY

The descriptive statistics above suggest that if any, the euro has had a differentiatedimpact on the two margins of trade. We use the decomposition of the intensive and theextensive margins with the bilateral dimension - as defined above - in our empiricalapproach17. Note that our analysis is constrained though by the fact that we only haveone year before the introduction of the euro. Our dataset also only contains exportflows from France. Our estimates therefore refer to the effect of the euro on Frenchexports to EZ partners, as compared to French exports to nonEZ partners. In ourestimations, the effect of the euro may therefore be under-estimated, since we neverrefer to export flows from nonEZ to nonEZ countries in our control group18.

17We therefore have no geographical component in the extensive margin of exports to each desti-nation country. The extensive margin is therefore defined as the number of varieties exported to eachpartner.

18Flam and Nordstrom (2003) indeed find some evidence of an effect of the euro on aggregate tradewhen the destination or the exporter in located outside the eurozone.

21


3.1. Empirical Strategy

We estimate two gravity equations with the intensive and extensive margins of Frenchexports as the dependent variable. We follow Flam and Nordstrom (2003) and firstinvestigate whether (i) there is an effect of euro on French exports beyond the elim-ination of the nominal exchange rate volatility, and (ii) this effect is changing overtime. We estimate the following two equations:

log(Ekjt) = α1EZ99 + α2EZ00 + α3EZ01 + α4EZ02 + α5EZ03 + α6V oljt

+α7log(rerjt)+α8log(rgdpjt)+α9log(distj)+α10contigj+α11CLj+κk+κt+εkjt

(1)

log(Ikjt) = β1EZ99 + β2EZ00 + β3EZ01 + β4EZ02 + β5EZ03 + β6V oljt

+β7log(rerjt)+β8log(rgdpjt)+β9log(distj)+β10contigj+β11CLj+κk+κt+µkjt

(2)

Where the EZyear dummies are interaction terms between an EZ dummy variable -indicating whether the destination country belongs to the eurozone, and the 1999 to2003 time dummies. Thus, for instance, EZ00 is equal to unity in 2000, if the desti-nation country is a member of the euro area, and zero otherwise. V oljt is a measureof the nominal exchange rate volatility between the two currencies19. rerjt is thereal exchange rate between France and the destination country j, at time t20; rgdpjt

is the real GDP in country j at time t; distj is the bilateral distance; contigj is adummy variable equal to unity if the two countries are contiguous; CLj is a dummyvariable equal to one if the destination market is French speaking. Finally, κk and κt

are respectively the industry and year fixed effects; εkjt and µkjt are the error terms.

We then complete our analysis by testing whether the introduction of the banknotesin 2002 had a positive effect on French export margins. The introduction of thebanknotes may have indeed improved competition within the eurozone by promotingprice transparency, which may have in turn resulted in further firms or products selec-tion on the export market. We test this effect with a simple difference-in-difference

19We come back to the construction of this variable in the data section. Since the euro dummies andthe volatility variable may be highly correlated, we also report estimation results without controllingfor volatility in the Table 9 in the Appendix Section.

20Thus, a higher rerjt means a real appreciation of the French currency against partner’s j currency.

22


approach, by introducing a dummy variable for exports to EZ countries over the2002-2003 period. We then estimate two gravity equations over the 1999-2003 pe-riod. Each year is thus treated by the initial effect of the euro, and we can isolate theeffect which is specific to the introduction in the banknotes.

log(Ekjt) = α1EZ02−03 +α2V oljt +α3log(rerjt)+α4log(rgdpjt)+α5log(distj)

+ α6contigj + α7CLj + α8EZ + κk + κt + εkjt (3)

log(Ikjt) = β1EZ02−03 + β2V oljt + β3log(rerjt) + β4log(rgdpjt) + β5log(distj)

+ β6contigj + β7CLj + β8EZ + κk + κt + µkjt (4)

Where EZ02−03 is a dummy variable which is equal unity for the period 2002-2003,if the trade partner is a member of the euro area - EZ, and zero otherwise. A positivecoefficient on the EZ02−03 dummy variable would mean that there has been a posi-tive effect of the introduction of the banknotes in 2002.

For each test related to the changing effect of the euro over time, or alternativelythe effect of the banknotes, we complete the empirical investigation by introducingan EU15 dummy variable - EU15j , in order to control for the fact that the effect ofthe euro on French exports is not driven by the belonging of the destination countryto the European Union. We also use a within fixed effect estimator that controls forindustry x country pair fixed effects in the estimation; this enables to control for re-verse causality. Indeed, countries that intensively trade together have an incentive toreduce transaction costs and uncertainty related to the nominal exchange rate volatil-ity. Consequently, part of the effect of the euro on trade may be due to this feature;controlling for country-pair fixed effects enables to treat this potential source of en-dogeneity21.

Finally, we propose some robustness checks at the end of the empirical section, bysuccessively dropping the nonEZworld and nonEZeurope groups from the controlgroup, as suggested by Baldwin (2006) and Flam and Nordstrom (2007). While there

21We report the estimation results with and without including the country pair fixed effects, since itdramatically reduces the number of degrees of freedom.

23


has been no further enlargement of the EU during the 1998-2003 period, EU policiesthat aim at driving a deeper integration within the EU may have led to a positive effecton intra-EU15 trade flows, that may be captured by the EZ dummies and lead to abiased estimate of the effect of the euro on French exports. Restricting the size of thecontrol group to countries that have benefited from those policies should thereforeenable to better capture the true effect of the euro.

3.2. Data

We use the trade data for individual French firms22 provided by the customs databasefor the period 1998-2003, and compute the French export margins - with a bilateraldimension - as described above. The data for bilateral distance, common language,and contiguity come from the CEPII. Data for real GDP come from the Penn WorldTables.

We compute the volatility of the exchange rate as Tenreyro (2007), by taking, foreach year in our sample, the standard deviation of the monthly variation of the nomi-nal exchange rate:

V oljt = Std.Dev.

(ejt,m − ejt,m−1

ejt,m−1

)With m = 1...12. V oljt is the yearly volatility of the monthly nominal exchange rateof the French currency against the foreign currency.

We compute our bilateral real exchange rate variable using the producer prices ofthe exporter - France - and importer countries, in the domestic currencies. The datafor producer price indices come from the International Financial Statistics (IMF) andthe OECD. Note that the producer price index for China is not available from thesesources; we use instead the PPI provided by the China Statistical Yearbook 2006(National Bureau of Statistics China). We also use bilateral nominal exchange ratesto compute our bilateral real exchange rate. All data related to nominal exchangerate come from the International Financial Statistics (IFS), and from the EuropeanCentral Bank.

After having introduced all these controls into the equations to be estimated, we ob-tain a balanced panel that consists of 50 partner countries for 6 years. Controlling for

22Within the EU, French customs collect information on the product (NC8 categories) exported byfirms when the annual cumulated value of all shipments of a firm (in the previous year) is above 100,000euros from 2001 onwards. This threshold was 99,100 euros in 2000 and 38,100 euros before. As regardsextra-eu exports, all shipments above 1,000 euros are reported.

24


the bilateral real exchange rate considerably reduces the number of importing coun-tries, due to data availability for the producer price indexes. However, we roughlycover 90 percent of the value of French exports throughout the considered period.We also have 25 ISIC industries in the final sample. We drop 3 industries - 314, 331and 354 - in order to obtain a balanced panel; the actual number of observations forthese three industries is indeed much reduced in the data. This leaves us with 7,500observations. We provide a list of all destination countries in Table 823.

4. ESTIMATION RESULTS

4.1. Main Results

We first estimate equations (1) and (2) to investigate the time-variant effect of theeuro, and report the results in Table 4. Using panel data with heterogenous individu-als requires to implement maximum likelihood techniques. In the first four columnsof Table 4, we implement a random effects GLS, and control for industry and timefixed effects. Columns III and IV additionally includes an EU15 membership dummyvariable, to control for the fact that a number of EU members may have adoptedthe euro because they actually intensively trade together. Finally, results reported incolumns V and VI are obtained using a - within - fixed effect estimator, which im-plies the use of industry x country-pair fixed effects, and additionally control for timefixed effects. This last methodology dramatically reduces the number of degrees offreedom, but enables to control for any source of reverse causality and auto-selectioninto the monetary union, as discussed in the previous section24.

Results first indicate that the variables that are traditionally used in the gravity equa-tion have the expected sign: bilateral distance has a negative effect on the number ofvarieties that are shipped to each industry, as well as on the average value of exportsby shipment. Contiguity has a positive effect on the two margins. French exportersexport a larger number of varieties to destination countries that share the same offi-cial language, but export a lower value of each variety to those countries. A possibleinterpretation is that sharing the same culture reduces the fixed cost to export, which

23The restriction to 50 destination countries may introduce some bias in the results since we removea large proportion of zeros from our dataset. Nevertheless, Helpman et al. (2007) show that most of thebias in estimating gravity equation comes from the absence of differentiation between the intensive andextensive margins, rather by the existence of zeros in bilateral trade matrices.

24Since volatility is close to zero for the eurozone, we control in a robustness check reported inAppendix Section, Table 9, that the coefficients on the EZ dummies remain unaffected when volatilityis excluded from the estimated equation.

25


Table 4: Trade effect of the euro, effect by year

I II III IV V VIDep. var. Extensive Intensive Extensive Intensive Extensive Intensive

Margin Margin Margin Margin Margin MarginEZ1999 0.102*** -0.035 0.060*** 0.002 0.051*** 0.018

(0.017) (0.029) (0.016) (0.03) (0.015) (0.03)EZ2000 0.061*** -0.042 0.019 -0.005 0.033** 0.037

(0.017) (0.031) (0.016) (0.032) (0.015) (0.031)EZ2001 0.049*** -0.064** 0.007 -0.027 0.011 0.003

(0.017) (0.029) (0.016) (0.03) (0.015) (0.029)EZ2002 0.075*** -0.016 0.034* 0.02 0.043** 0.055

(0.021) (0.035) (0.02) (0.037) (0.02) (0.036)EZ2003 0.221*** -0.076** 0.178*** -0.039 0.189*** -0.004

(0.022) (0.038) (0.021) (0.039) (0.022) (0.04)V olatilityjt -2.055*** -1.376*** -2.013*** -1.434*** -1.324*** -0.708

(0.152) (0.418) (0.153) (0.424) (0.168) (0.439)RERjt -0.088*** -0.018 -0.083*** -0.019 -0.351*** -0.340**

(0.012) (0.013) (0.012) (0.013) (0.065) (0.155)RGDPjt 0.587*** 0.355*** 0.537*** 0.363*** 1.328*** 1.191***

(0.017) (0.02) (0.017) (0.021) (0.086) (0.266)DISTj -0.716*** -0.217*** -0.513*** -0.248***

(0.022) (0.023) (0.027) (0.029)Contigj 0.588*** -0.09 0.307*** 0.125**

(0.069) (0.065) (0.067) (0.061)CLj 0.598*** -0.04 0.938*** -0.090*

(0.062) (0.05) (0.066) (0.052)EU15j 0.928*** -0.160***

-0.058 -0.054Nb observations 7350 7350 7350 7350 7350 7350Estimation Method RE RE RE RE FE FEFixed effects Industry Industry Industry Industry Ind x pair Ind x pair

& year & year & year & year & year & year

Note: Significance levels: *10%, **5%, ***1%. All variables - with the exception of dummy variables- are in logarithms. Robust standard errors in parentheses. RE stands for random effects GLS, FE standsfor within fixed effect estimation.

26


enables to export new varieties but reduces the average value by shipment, by mak-ing smaller exporters to enter. For these three variables, the marginal effect is largeron the extensive margin, which may imply that bilateral distance, contiguity and theexistence of a common language affect the fixed entry cost, and therefore the produc-tivity cutoff that enables a firm to export a new variety. Finally, a real appreciationof the domestic currency in France has always a negative effect, at least on the exten-sive margin, while the real GDP of the destination country positively influences thetwo margins. Note that all results related to gravity variables are very consistent withthe estimation results presented in Mayer and Ottaviano (2007), that use the samedatabase.

Estimation results reported in column I point to a positive effect of the euro on theextensive margin of French exports, for each year consecutive to 1998. The positiveeffect however decreases until 2001 and then increases again until 2003. The Waldtest on coefficients, reported in Table 10 in the Appendix Section, cannot reject theequality of coefficients on EZ2000, EZ2001 and EZ2002 variables for the first column.While this effect may be associated to the evolution of the euro parity with regards toother major currencies over the period - with a weak euro in 2001 generating a tradediversion effect, results remain unchanged when we control for real exchange ratemovements25. The introduction of the EU15 dummy variable in column III does notqualitatively modify the results, even though it severely reduces the positive impactof the euro on the extensive margin. Finally, results from the estimation introducingindustry x country-pair fixed effects reported in column V are consistent with thosein columns I and III, suggesting a positive and exogenous effect of the euro on theextensive margin. Results on the intensive margin are also very clear: all coefficientson the euro time dummies are non-significant, or even negative in 2001 and 2003, incolumn II.

Results related to the volatility of the nominal exchange rate suggest a negative ef-fect on both the intensive and extensive margins, with the exception of estimationsusing a within estimator in columns V and VI. We report estimation results withoutcontrolling for volatility in Table 9 of the Appendix Section. Results indicate that thecoefficients on the euro dummies remain unaffected.

Overall, results suggest that Currency Unions may benefit to exports through theelimination of nominal exchange rate volatility, but also through reduced trade costs

25Note that, as suggested previously, the evolution in the reporting thresholds imposed to exportingfirms by the French customs may have an influence on our results, by statistically restricting the numberof exporting firms to the EU15, especially by 2000.

27


and increased price transparency. While the effect related to nominal exchange ratevolatility translates both into the number of varieties and the average value of ship-ments by variety, as suggested by Baldwin and Taglioni (2004), we find some evi-dence of an additional effect of the euro on trade that fully translates into the numberof varieties that are traded. This result is consistent with a decrease in the fixed cost oftrading across borders. It is also consistent with an increased competition within theeuro area related to price transparency, generating new entries into the export market,with the form of new varieties. Note that the descriptive statistics suggest that theeffect should be related to existing firms exporting more products to more partners,contributing positively to the extensive margin, while some exporting firms disappearduring the period. Our findings can therefore also be related to the model developedby Bernard et al. (2006), who find that part of the adjustment consecutive to tradeliberalization and a variation in trade costs operate on the number of goods exportedby each firm.

We take the coefficients on the EZ2003 variable in the estimation of the two equa-tions with the EU15 control - in column III - for our quantification exercise. We findthat the difference between the number of varieties exported to EZ partners and non-EZ partners, has increased by 19.4% between 1998 and 2003.26.

We then test the impact of the introduction of the banknotes in 2002, and estimateequations (3) and (4) for the period 1999-2003 and report the results in Table 5. Theresults indicate that the coefficients on contiguity and common language are positivefor the extensive margin, while the common language variable has a negative effecton the intensive margin. These results therefore remain extremely consistent withthose presented in Table 4, with a larger effect of distance, contiguity and commonlanguage on the extensive margin. The coefficient on distance is always negative, aswell as the coefficient on the real exchange rate - but only for the extensive margin,while the effect of the real GDP in the destination country is always positive.

Most importantly, the coefficient on the EZ2002−2003 dummy variable in the Tableis positive and highly significant for the extensive margin, and not significant for theintensive margin. This suggests that the increase in the price transparency relatedto the introduction of the banknotes has led to further firm or product selection onthe export market. Results remain extremely stable when we introduce the EU15control, and when we introduce the industry x country-pair fixed effects in columnsV and VI of Table 5. This implies that the difference between the number of varieties

26The marginal effect is computed as follows: e0.178 − 1 = 0.194

28


Table 5: Trade effect of the euro, period 2002-2003.


Margin Margin Margin Margin Margin MarginBanknotes 0.076*** -0.004 0.075*** -0.004 0.067*** 0.011

(-0.013) (-0.024) (-0.013) (-0.024) (-0.014) (-0.027)V olatilityjt -2.167*** -1.947*** -2.179*** -1.951*** -1.405*** -2.211***

(-0.179) (-0.569) (-0.18) (-0.571) (-0.195) (-0.564)RERjt -0.098*** -0.018 -0.089*** -0.019 -0.662*** -0.03

(-0.012) (-0.013) (-0.012) (-0.013) (-0.078) (-0.02)RGDPjt 0.554*** 0.360*** 0.522*** 0.361*** 1.184*** 0.332***

(-0.017) (-0.02) (-0.018) (-0.022) (-0.108) (-0.026)DISTj -0.596*** -0.245*** -0.513*** -0.250***

(-0.025) (-0.026) (-0.027) (-0.031)Contigj 0.231*** 0.166** 0.304*** 0.162**

(-0.075) (-0.067) (-0.074) (-0.069)CLj 0.824*** -0.076 0.928*** -0.082

(-0.067) (-0.053 (-0.067) (-0.054)EZj 0.835*** -0.225*** 0.133 -0.188***

(-0.059) (-0.049) (-0.088) (-0.07)EU15j 0.848*** -0.045

(-0.091) (-0.08)Nb observations 6125 6125 6125 6125 6125 6125Estimation Method RE RE RE RE FE FEFixed effects Industry Industry Industry Industry Ind x pair Ind x pair



exported to EZ partners and nonEZ partners has increased by about 7.8% betweenthe 1999-2001 and 2002-2003 periods27. This effect should be interpreted as theeffect of the euro on French exports, that is specifically related to the introduction ofthe banknotes. As in the previous Table, this result is independent of the effect ofreduced volatility, but also independent from the initial effect of the euro on Frenchexports in 1999, since we only keep the 1999-2002 period for the above estimations.

27We take the coefficient on the banknotes dummy in column III: exp(0.075)-1=0.078

29


4.2. Robustness

In this last exercise, we refine the analysis by restricting the size of the control group.This methodology enables to implicitly control for the fact that some time varianteffect may affect the EZ countries as well as neighboring countries in the same way.In particular, EU policies that aimed at promoting the integration within Europe mayintroduce some bias in our results. We therefore successively drop from the controlgroup the nonEZworld and nonEZeurope destination regions. Note that for this exer-cise, we keep the simplest form of the equation to be estimated, and only keep the realGDP of the destination country as well as bilateral distance as controls28. We firstprovide the results of the robustness checks for the time effect of the euro in Table6. Results indicate that most of the effect of the euro on the extensive margin disap-pears when we remove nonEZworld from the control group - only the 1999 and 2003time dummies of the euro have a positive and significant coefficient, while the eurodoes not seem to have had a significant differentiated effect on the extensive marginwhen we only keep nonEZeu as the control group. These results are very consistentwith Flam and Nordstrom (2007)29, that indeed find that the effect of the euro on theextensive margin - defined by the number of product categories exported to a givenpartner - indeed disappears, when one restricts the sample of importing countries toEZ and nonEZeu countries. The effect of the euro on the intensive margin remainsunaffected though.

We then test whether our result related to the introduction of the banknotes is stableto the restriction of the sample size. We report the results of estimations in Table 7.Results indicate that removing nonEZworld or nonEZworld and nonEZeurope fromthe control group leads to a non-significant differentiated effect of the introduction ofthe banknotes on the extensive margin.

All these results therefore indicate that the positive effect of the euro introductionon the extensive margin of French exports is influenced by the size of the controlgroup. While we indeed find a positive effect of the euro introduction on exports toEZ partners, as compared to nonEZworld partners, the effect is ambiguous when weestimate the effect by only keeping European countries in the control group. Theseresults could first be interpreted by the fact that the estimated positive effect of the

28The real exchange rate, volatility, common language and contiguity variables are indeed extremelyaffected by the size of the sample of destination countries, since we only have France as the exporter.Keeping those controls for our robustness exercise would therefore lead to biased results.

29See Table 2 of their paper.

30


Table 6: Time effect of the euro, restricting the control group

I II III IV V VIControl group all nonEZ nonEZeu, nonEZeurope nonEZeuDep. var. Extensive Intensive Extensive Intensive Extensive Intensive

Margin Margin Margin Margin Margin MarginEZ1999 0.075*** -0.033 0.061*** -0.032 0.021 -0.034**

(0.017) (0.029) (0.022) (0.038) (0.026) (0.017)EZ2000 0.036** -0.036 0.01 -0.029 0.016 0.031

(0.016) (0.031) (0.022) (0.039) (0.024) (0.02)EZ2001 0.02 -0.064** -0.026 -0.082** 0.011 0.032

(0.017) (0.029) (0.022) (0.039) (0.027) (0.02)EZ2002 0.071*** -0.001 -0.035 -0.07 0.021 -0.045**

(0.021) (0.033) (0.027) (0.044) (0.038) (0.022)EZ2003 0.208*** -0.068* 0.054* -0.130** 0.039 -0.110***

(0.023) (0.036) (0.03) (0.054) (0.042) (0.025)RGDPjt 0.670*** 0.369*** 0.770*** 0.354*** 0.372*** 0.376***

(0.017) (0.019) (0.023) (0.022) (0.03) (0.065)DISTj -0.883*** -0.248*** -1.054*** -0.364*** -1.000*** -0.373*

(0.02) (0.019) (0.069) (0.057) (0.084) (0.212)Nb Observations 7350 7350 3000 3000 1650 1650Estimation Method RE RE RE RE RE REFixed effects Industry Industry Industry Industry Industry Industry


Note: Significance levels: *10%, **5%, ***1%. All variables - with the exception of dummy variables- are in logarithms. Robust standard errors in parentheses. RE stands for random effects GLS.

euro actually captures the positive effect of an overall European integration. Indeed,even if there has been no further enlargement during the 1998-2003 period, policiesthat aimed at promoting a deeper integration within the EU may have led to a posi-tive effect on the extensive margin of French exports, which would lead to a biasedestimated effect of the euro on trade.

Flam and Nordstrom (2007) and Baldwin (2006) also suggest that there is a posi-tive effect of the euro for trade from EZ to nonEZ countries. This effect can thereforepotentially explain why the effect of the euro on the extensive margin progressivelydisappears when we restrict the size of the control group. Indeed, if the euro had apositive effect on exports between EZ and nonEZ countries, this may especially haveaffected exports to nonEZ neighboring countries. Baldwin (2006) indeed suggeststhat the effect of the euro on exports should be the highest when trade costs are alsolower.

31


Table 7: Effect of the banknotes, restricting the control group

I II III IV V VIControl group all nonEZ nonEZeu, nonEZeurope nonEZeuDep. Var. Extensive Intensive Extensive Intensive Extensive Intensive

Margin Margin Margin Margin Margin MarginBanknotes 0.095*** 0.01 -0.01 -0.051* 0.014 0.026

(-0.013) (-0.025) (-0.017) (-0.031) (-0.024) (-0.039)RGDPjt 0.611*** 0.379*** 0.651*** 0.379*** 0.385*** 0.463***

(-0.018) (-0.02) (-0.023) (-0.021) (-0.031) (-0.025)DISTj -0.742*** -0.273*** -1.014*** -0.359*** -0.956*** -0.267***

(-0.025) (-0.025) (-0.069) (-0.06) (-0.087) (-0.069)EZj 0.763*** -0.141*** 0.671*** -0.143*** 0.184*** -0.085

(-0.057) (-0.044) (-0.049) (-0.04) (-0.067) (-0.066)Nb observations 6250 6250 2625 2625 1375 1375Estimation Method RE RE RE RE RE REFixed effects Industry Industry Industry Industry Industry Industry


Note: Significance levels: *10%, **5%, ***1%. All variables - with the exception of dummy variables- are in logarithms. Robust standard errors in parentheses. RE stands for random effects GLS.

Two last limitations are associated with the peculiar nature of our database. Firstly,we only have one year before the introduction of the euro. Secondly, we cannot con-trol for export flows between nonEZ to nonEZ countries, since we only have Franceas the exporter. This may therefore reduce our estimated effect of the euro on theintensive and extensive margins. Nevertheless, our results clearly suggest that theeffect of the euro on French exports, if any, translates into the extensive margin.

5. CONCLUSION

This paper contributes to the recent literature that aims at measuring the effect of theadoption of the euro on trade. Using French firm-level data, we show that the effectof the euro mainly translates into the extensive margin, i.e. the number of varietiesthat are exported to euro area members, while the intensive margin, i.e. the averagevalue of shipments by variety, does not seem to be affected. According to our esti-mations, the difference between the number of varieties exported to EZ partners andnon-EZ partners, has increased by 19.4% between 1998 and 2003; we also estimatethe effect specific to the introduction of the banknotes at 7.8%.

32


Those results are very consistent with the view that reduced fixed costs enables toexport new varieties, as suggested by Melitz (2003), while increased transparencymay have generated a larger competition within the eurozone, therefore contributingto the selection of firms. Nevertheless, the descriptive statistics suggest that the effecttranslated into more products exported by each firm rather than by an increase in thenumber of exporting firms. Our contribution can therefore be related to the modelof multi-product firms proposed by Bernard et al. (2006). We also find that nomi-nal exchange rate volatility has a negative effect on both the extensive and intensivemargins, which is also consistent with the theoretical approach proposed by Baldwinand Taglioni (2004). Finally we find that results are conditioned to the choice of thecontrol group, the effect of the euro indeed disappears when we restrict the controlgroup to EU15 non-eurozone member countries. A pessimistic interpretation of thislast result would conclude that the effect captured by the euro dummies actually re-flect other reduction of trade costs related to European integration as a whole. Anoptimistic view would conclude to a ”spillover effect” of the euro on neighboringcountries, as suggested by Baldwin (2006).

Overall, this paper considerably improves the estimation of the micro effects of euroadoption on trade, by making use of a unique database on firm-level exports.

33


REFERENCES

Baldwin, R. and Taglioni, D. (2004). Positive oca criteria: Microfoundations for therose effect. COE/RES Discussion Paper Series No. 34.

Baldwin, R. E. (2006). The euro’s trade effect. ECB Working Paper N. 594.

Baldwin, R. E. and Nino, V. D. (2006). Euros and zeros: The common currencyeffect on trade in new goods. NBER Working Papers N. 12673.

Bernard, A. B., Jensen, J. B., Redding, S. J., and Schott, P. K. (2007). Firms ininternational trade. Journal of Economic Perspectives, 21(3):105–130.

Bernard, A. B., Redding, S. J., and Schott, P. K. (2006). Multi-product firms andtrade liberalization. NBER Working Papers n. 12782.

Flam, H. and Nordstrom, H. (2003). Trade volume effects of the euro: Aggregateand sector estimates. IIES Seminar Paper No. 746.

Flam, H. and Nordstrom, H. (2007). Explaining large euro effects on trade: theextensive margin and vertical specialization. Manuscript.

Helpman, E., Melitz, M., and Rubinstein, Y. (2007). Estimating trade flows: Tradingpartners and trading volumes. NBER Working Paper N.12927.

Mayer, T. and Ottaviano, G. (2007). The happy few: the internationalization ofeuropean firms. Bruegel Blueprint Series vol. 3, Brussels.

Melitz, M. J. (2003). The impact of trade on intra-industry reallocations and aggre-gate industry productivity. Econometrica, 71(6):1695–1725.

Rose, A. K. (2000). One money, one market: the effect of common currencies ontrade. Economic Policy, 15(30):7–46.

Tenreyro, S. (2007). On the trade impact of nominal exchange rate volatility. Journalof Development Economics, 82(2):485–508.

34


APPENDIX

Table 8: List of Destination Countries

Argentina JapanAustralia KoreaAustria LithuaniaBelgium and Luxembourg LatviaBrazil MexicoCanada MalaysiaSwitzerland NetherlandsChile NorwayChina New ZealandColumbia PeruCzech Republic PhilippinesGermany PolandAlgeria PortugalEgypt SingaporeSpain El SalvadorEstonia SlovakiaFinland SloveniaUnited Kingdom SwedenHong Kong ThailandCroatia TunisiaHungary TurkeyIndonesia UruguayIndia United StatesIreland VenezuelaItaly South Africa

35


Table 9: Trade effect of the euro, effect by year. No control for volatility


Margin Margin Margin Margin Margin MarginEZ1999 0.084*** -0.042 0.040** -0.009 0.040*** 0.012

(0.017) (0.029) (0.016) (0.03) (0.015) (0.03)EZ2000 0.048*** -0.044 0.002 -0.011 0.033** 0.039

(0.017) (0.031) (0.016) (0.032) (0.015) (0.03)EZ2001 0.031* -0.072** -0.013 -0.039 0.004 -0.002

(0.017) (0.029) (0.016) (0.03) (0.015) (0.029)EZ2002 0.079*** -0.01 0.034* 0.023 0.046** 0.055

(0.022) (0.035) (0.02) (0.037) (0.02) (0.036)EZ2003 0.209*** -0.078** 0.163*** -0.046 0.172*** -0.013

(0.023) (0.037) (0.022) (0.038) (0.022) (0.04)RERjt -0.119*** -0.022* -0.110*** -0.023* -0.634*** -0.493***

(0.013) (0.012) (0.012) (0.012) (0.059) (0.133)RGDPjt 0.629*** 0.364*** 0.569*** 0.372*** 1.460*** 1.261***

(0.018) (0.021) (0.018) (0.022) (0.085) (0.275)DISTj -0.728*** -0.239*** -0.517*** -0.266***

(0.023) (0.024) (0.028) (0.03)Contigj 0.665*** 0.108* 0.348*** 0.142**

(0.071) (0.063) (0.068) (0.06)CLj 0.552*** -0.085* 0.908*** -0.129**

(0.062) (0.048) (0.065) (0.051)EU15j 0.996*** -0.140***

(0.058) (0.051)Nb observations 7350 7350 7350 7350 7350 7350Estimation Method RE RE RE RE FE FEFixed effects Industry Industry Industry Industry Ind x pair Ind x pair



36


Tabl

e10

:Wal

dte

stfo

requ

ality

ofpa

ram

eter

son

EZ

time

dum

mie

s

Test

onpa

ram

eter

sfr

omTa

ble

4,co

lum

nI

Var

iabl

esC

hi2

Pro

b>

chi2

Equ

ality

EZ

2001

&E

Z1999

11.5

30.

001

No

EZ

2001

&E

Z2000

0.60

0.43

8Y

esE

Z2001

&E

Z2002

1.82

0.17

7Y

esE

Z2001

&E

Z2003

66.2

40.

000

No

37

CEPII, Working Paper No 2008 – 06

38

LIST OF WORKING PAPERS RELEASED BY CEPII1

No Title Authors

2008-05 On the Influence of Oil Prices on Economic Activityand Other Macroeconomic and Financial Variables

F. Lescaroux& V. Mignon

2008-04 An Impact Study of the EU-ACP EconomicPartnership Agreements (EPAs) in the Six ACPRegions

L. Fontagné,D. Laborde

& C. Mitaritonna

2008-03 The Brave New World of Cross-Regionalism A. Tovias

2008-02 Equilibrium Exchange Rates: a Guidebook for theEuro-Dollar Rate

A. Bénassy-Quéré,S. Béreau

& V. Mignon

2008-01 How Robust are Estimated Equilibrium ExchangeRates? A Panel BEER Approach

A. Bénassy-Quéré,S. Béreau

& V. Mignon

2007-24 Testing the Finance-Growth Link: Is there aDifference between Developed and DevelopingCountries?

G. Dufrénot,V. Mignon

& A. Péguin-Feissolle

2007-23 Labor Migration: Macoeconomic and Demographicoutlook for Europe and Neighborhood Regions

V. Borgy& X. Chojnicki

2007-22 Economic Geography, Spatial Dependence andIncome Inequality in China

L. Hering& S. Poncet

2007-21 Does FDI in Manufacturing Cause FDI in BusinessServices? Evidence from French Firm-Level Data

B. Nefussi& C. Schwellnus

2007-20 Bilateral Trade of Cultural Goods A.C. Disdier,S.H.T. Tai,

L. Fontagné& T. Mayer

2007-19 China and India in International Trade: fromLaggards to Leaders?

F. Lemoine& D. Ünal-Kesenci

2007-18 How Remote is the Offshoring Threat K. Head, T. Mayer& J. Ries

1 Working papers are circulated free of charge as far as stocks are available; thank you to send your request

to CEPII, Sylvie Hurion, 9, rue Georges-Pitard, 75015 Paris, or by fax : (33) 01 53 68 55 04 or by [email protected]. Also available on: \\www.cepii.fr. Working papers with * are out of print. They cannevertheless be consulted and downloaded from this website.


39

2007-17 Costs and Benefits of Euro Membership: aCounterfactual Analysis

E. Dubois,J. Héricourt

& V. Mignon

2007-16 Location Decisions and Minimum Wages I. Méjean& L. Patureau

2007-15 MIRAGE, Updated Version of the Model for TradePolicy Analysis Focus on Agriculture and Dynamics

Y. Decreux& H. Valin

2007-14 Mondialisation des services de la mesure à l'analyse I. Bendisoun& D. Ünal-Kesenci

2007-13 How are wages set in Beijing? J. De Sousa& S. Poncet

2007-12 IMF Quotas at Year 2030 A. Bénassy-Quéré,S. Béreau,

Y. Decreux, C. Gouel& S. Poncet

2007-11 FDI and Credit Constraints: Firm Level Evidence inChina

J. Héricourt& S. Poncet

2007-10 Fiscal Policy in Real Time J. Cimadomo

2007-09 Global Ageing and Macroeconomic Consequences ofDemographic Uncertainty in a Multi-regional Model

J. Alho & V. Borgy

2007-08 The Effect of Domestic Regulation on Services TradeRevisited

C. Schwellnus

2007-07 The Location of Domestic and Foreign ProductionAffiliates by French Multinational Firms

T. Mayer, I. Méjean & B. Nefussi

2007-06 Specialisation across Varieties within Products andNorth-South Competition

L. Fontagné,G. Gaulier

& S. Zignago

2007-05 Trade Costs and the Home Market Effect M. Crozet& F. Trionfetti

2007-04 The Impact of Regulations on Agricultural Trade:Evidence from SPS and TBT Agreements

A.-C. Disdier,L. Fontagné

& M. Mimouni

2007-03 International Comparisons of Living Standards byEquivalent Incomes

M. Fleurbaey &G. Gaulier

2007-02 Does Risk Aversion Drive Financial Crises? Testingthe Predictive Power of Empirical Indicators

V. Coudert & M. Gex

CEPIIDOCUMENTS DE TRAVAIL / WORKING PAPERS

Si vous souhaitez recevoir des Documents de travail,merci de remplir le coupon-réponse ci-joint et de le retourner à :

Should you wish to receive copies of the CEPII’s Working papers,just fill the reply card and return it to:

Sylvie HURION – PublicationsCEPII – 9, rue Georges-Pitard – 75740 Paris – Fax : (33) 1.53.68.55.04

[email protected]

M./Mme / Mr./Mrs ..................................................................................................................

Nom-Prénom / Name-First name .............................................................................................

Titre / Title ...............................................................................................................................

Service / Department................................................................................................................

Organisme / Organisation ........................................................................................................

Adresse / Address.....................................................................................................................

Ville & CP / City & post code..................................................................................................Pays / Country.............................................................. Tél......................................................Your e-mail ..............................................................................................................................

Désire recevoir les Document de travail du CEPII n° :

Wish to receive the CEPII’s Working Papers No:..................................................................

..................................................................................................................................................

..................................................................................................................................................

..................................................................................................................................................

..................................................................................................................................................

..................................................................................................................................................

..................................................................................................................................................

..................................................................................................................................................

..................................................................................................................................................

Souhaite être placé sur la liste de diffusion permanente (pour les bibliothèques)Wish to be placed on the standing mailing list (for Libraries).

Date post:	25-Nov-2021
Category:	Documents
Upload:	others
View:	5 times
Download:	0 times